LLM Stacks

Structured repository to docs, techniques, patterns, experiments, and even some prompt about / for Large Language Models (LLMs) - Software Development Oriented

What you will find here

• Prompting templates – reusable prompt skeletons designed for chats, code generation, and evaluation.
• Techniques – step-by-step guides that explain how to combine LLMs and humans effectively (TDD, Multi-Agent, Live-RAG …).
• Stacks – opinionated selections of libraries and tools for tasks such as data analysis, RAG, or UI prototyping.
• Model Skills – a subjective assessment based on hands-on use of several leading models.

Getting started

1. Clone the repo and create a Python virtual environment (optional, but usefull in general if you haven’t try it already).
2. Explore the folders above to find ready-to-use templates, stacks, or techniques relevant to your project.

.
├── prompting/              # Prompt templates and experiments
│   └── templates/          # Parameterised prompt templates
├── stacks/                 # Technology stacks for different problem domains
├── techniques/             # Methodologies for working with LLMs (TDD, Multi-Agent, Live-RAG, …)
│   ├── Live-RAG/
│   ├── MultiAgent/
│   └── TestDrivenDevelopment/
└── LICENSE                 # MIT license

3. Contribute back! Feel free to open pull-requests with new templates, methodologies, or improvements.

Other Tools & Related Repos

• RAG-openai-chats: a step-by-step guide on how to build a RAG system using OpenAI's Exported Chats.
• Local-RAG-Prototype: Local RAG prototype with a ports & adapted architecure, ready to plug in / apply custom changes!
• RepoGPT: code repository analysis tool. Creates summaries for largue repos, in different formats / structures. Usefull as code-structure compressor for LLMs. Support jq syntax + filters / abstraction levels for advance queries.
• Ðata Analysis and OSINT: an example on how to use efficiently LLMs on a research project - OSINT, data analysis, data scraping.
• MCP-local-agent: a local example for an MCP based agent - added custom tools to a Gemini3:4b model used via Ollama local API.
• Open DeepResearch: Open source alternative to Deep Research. [Gemini Oriented]
• Real Prompt Engineering: Elder-Plinius - L1B3RT4S, CL4R1T4S..

Roadmap

• Add unit & integration test examples for each technique.
• Expand the stacks section with containerised starter projects.
• Provide reference implementations for automated evaluation pipelines.
• Feel free to propse ideas, techniques, or tools. Do you miss something? Let us know!

License

This project is licensed under the MIT License – see the LICENSE file for details.

---

Feel free to raise issues if you spot inconsistencies or have suggestions. Happy hacking! 🚀